koboldcpp v1.118.1 fixes multi-GPU issues and adds Qwen3TTS support

The latest release of koboldcpp (v1.118.1) addresses several critical bugs, including a fix for incorrect GPU selection in multi-GPU setups which could lead to performance degradation or service interruptions. Additionally, the update supports qwen3tts aliases for all voices and aligns RPC behavior with llama.cpp, enhancing compatibility between different AI services. Operators running koboldcpp on systems with multiple GPUs should upgrade immediately to ensure correct resource allocation and avoid potential disruptions.